articles

Involvement of chemokine receptors in

breast cancer metastasis
Anja MuÈller*²³, Bernhard Homey*²³, Hortensia Soto*, Nianfeng Ge*, Daniel Catron*, Matthew E. Buchanan*, Terri McClanahan*,
Erin Murphy*, Wei Yuan*, Stephan N. Wagner§, Jose Luis Barrerak, Alejandro Mohark¶, Emma VeraÂsteguik & Albert Zlotnik*
* Department of Immunology, DNAX Research Institute, 901 California Avenue, Palo Alto, California 94304, USA
² Departments of Radiation Oncology and Dermatology, Heinrich-Heine University, Moorenstrasse 5, D-40225 DuÈsseldorf, Germany
§ Department of Dermatology, University of Essen, Hufelandstrasse 55, D-45147 Essen, Germany
k Instituto Nacional de CancerologõÂa, Av. San Fernando 22, Tlalpan 14000 D.F., MeÂxico
¶ Instituto de Investigaciones BiomeÂdicas, Ciudad Universitaria, 04510 D.F., MeÂxico
³ These authors contributed equally to this work
............................................................................................................................................................................................................................................................................

Breast cancer is characterized by a distinct metastatic pattern involving the regional lymph nodes, bone marrow, lung and liver.
Tumour cell migration and metastasis share many similarities with leukocyte traf®cking, which is critically regulated by
chemokines and their receptors. Here we report that the chemokine receptors CXCR4 and CCR7 are highly expressed in human
breast cancer cells, malignant breast tumours and metastases. Their respective ligands CXCL12/SDF-1a and CCL21/6Ckine
exhibit peak levels of expression in organs representing the ®rst destinations of breast cancer metastasis. In breast cancer cells,
signalling through CXCR4 or CCR7 mediates actin polymerization and pseudopodia formation, and subsequently induces
chemotactic and invasive responses. In vivo, neutralizing the interactions of CXCL12/CXCR4 signi®cantly impairs metastasis of
breast cancer cells to regional lymph nodes and lung. Malignant melanoma, which has a similar metastatic pattern as breast
cancer but also a high incidence of skin metastases, shows high expression levels of CCR10 in addition to CXCR4 and CCR7. Our
®ndings indicate that chemokines and their receptors have a critical role in determining the metastatic destination of tumour cells.

Metastasis is the result of several sequential steps and represents a Three different patterns of receptor expression were observed.
highly organized, non-random and organ-selective process1. First, the chemokine receptor was expressed in normal mammary
Although a number of molecules have been implicated in the epithelial cells, but uniformly downregulated by breast cancer cells
metastasis of breast cancer, the precise mechanisms determining (for example, CXCR2). Second, both normal mammary epithelial
the directional migration and invasion of tumour cells into speci®c cells and malignant cells expressed signi®cant levels of receptor
organs remain to be established1±3. mRNA; this pattern was detected for CCR7 and CCR8, with CCR7
Chemokines are a superfamily of small, cytokine-like proteins showing the most consistent upregulation in breast cancer cells.
that induce, through their interaction with G-protein-coupled Third, receptor expression was undetectable in normal mammary
receptors, cytoskeletal rearrangement, ®rm adhesion to endothelial epithelial cells but markedly upregulated in breast cancer cells (for
cells and directional migration4±6. These secreted proteins act in a example, CXCR4).
coordinated fashion with cell-surface proteins, including integrins, Flow cytometric analyses con®rmed strong cell-surface expres-
to direct the speci®c homing of various subsets of haematopoietic sion of CXCR4 on breast cancer cell lines (data not shown), and on
cells to speci®c anatomical sites4±10. primary breast cancer cells (82.33±97.98% CXCR4-positive malig-
We therefore thought that tumour cells may use chemokine- nant cells) recovered from patients with malignant pleural effusion
mediated mechanisms such as those regulating leukocyte traf®cking (n = 5) (see Supplementary Information).
during the process of metastasis. Here we report that tumour cells We quantitatively measured CCR7 and CXCR4 mRNA expres-
express a distinct, non-random pattern of functionally active sion in primary tumours of human invasive lobular or ductal breast
chemokine receptors. Signalling through CXCR4 or CCR7 mediates carcinoma (n = 12), and in normal mammary gland (n = 5) and
actin polymerization and pseudopodia formation in breast cancer its cellular constituents (Fig. 2a, b). According to the TNM
cells, and induces chemotactic and invasive responses. In addition, classi®cation11, six tumours were classi®ed as T2, four as T3 and
we ®nd that organs representing the main sites of breast cancer two as T4, with evidence of axillary nodal metastasis in ®ve cases.
metastasis are the most abundant sources of ligands for these Figure 2a and b shows that mRNA of both CCR7 and CXCR4 was
tumour-associated receptors. In vivo, neutralizing the interactions signi®cantly upregulated in primary breast tumours compared with
of CXCL12/CXCR4 leads to a signi®cant inhibition of lymph-node normal mammary gland tissue (P , 0.05, P , 0.005, respectively).
and lung metastasis. Normal primary mammary epithelial and stromal cells derived
from three different donors did not express detectable levels of
Expression of chemokine receptors in tumour cells CXCR4 mRNA. There was no signi®cant correlation between the
To determine whether chemokine/chemokine receptor interactions TNM status of the 12 patients studied and absolute levels of CCR7
are involved in the metastatic process, we performed a comprehen- and CXCR4 mRNA expression in primary tumours.
sive, quantitative analysis of the expression of all known chemokine We con®rmed in vivo protein expression of CXCR4 by immuno-
receptors (CCR1±CCR10, CXCR1±CXCR5, XCR1 and CX3CR1) in histochemistry (Fig. 2c±n). Stained tissue sections from normal
seven human breast cancer cell lines, and compared expression in mammary gland (n = 4) (Fig. 2c±e) or invasive ductal carcinoma
normal primary mammary epithelial cells. Absolute messenger (n = 8) (Fig. 2f±k) indicated that breast cancer cells express high
RNA levels determined using real-time quantitative polymerase levels of CXCR4, but normal mammary ductal cells do not express
chain reaction (PCR) showed that breast cancer cells express this receptor. In addition to primary tumours, axillary lymph-node
chemokine receptors in a de®ned rather than a random manner metastases (n = 5) (Fig. 2l±n) and distant metastases of lung and
(Fig. 1a). liver (n = 8) (data not shown) exhibited strong CXCR4 expression

50 © 2001 Macmillan Magazines Ltd NATURE | VOL 410 | 1 MARCH 2001 | www.nature.com
 articles
in tumour cells. showed most abundant expression in lymph nodes; however,
Although CXCL12/SDF-1 mRNA has been reported to be CCL21 was expressed at higher levels, supporting its critical role
expressed in many different tissues12, quantitative analysis of in the homing of cells into lymph nodes14,15 (Fig. 2p; and see
CXCL12 expression in many human organs revealed that CXCL12 Supplementary Information). Comparing CXCL12 and CCL21
mRNA is expressed preferentially in lymph nodes, lung, liver and expression in primary breast tumours and breast cancer cell lines
bone marrow, and shows markedly lower expression in small with that in normal human organs showed that tumour cells express
intestine, kidney, skin, brain and skeletal muscle (Fig. 2o). Organs low or undetectable amounts of mRNA of these particular chemo-
exhibiting the highest CXCL12 expression represent the most kines (see Supplementary Information).
common sites of metastasis in breast cancer13. To determine whether the distinct expression of chemokine
In addition, both CCL21/6Ckine and CCL19/MIP-3b had receptors is unique to breast cancer, we also analysed malignant
melanoma, which, like breast cancer, preferentially develops lymph-
node, lung, liver and bone-marrow metastases, but also has a high
a CX3CR1 frequency of skin metastases. In addition to CXCR4 and CCR7,
malignant melanoma cells expressed high levels of CCR10 mRNA as
XCR1
compared with normal primary melanocytes (Fig. 1b). Notably, the
CXCR5 CCR10 ligand CCL27/CTACK represents a skin-speci®c homeo-
CXCR4 static chemokine (Fig. 2q) that has been associated with the homing
CXCR3 of memory T cells into the skin7,8.
CXCR2 Mammary
epithelial cells Breast cancer cells express active CXCR4 and CCR7
CXCR1 Hs5787T In vitro, chemokine ligand±receptor interactions trigger intracellu-
CCR10 T-47D lar actin polymerization in leukocytes, a process that is prerequisite
CCR9
DU-4475 for cell motility and migration16,17. Consistent with ®ndings in
MDA-MB-361 leukocytes16, CXCL12 (100 nM) and CCL21 (100 nM) induced,
CCR8
MDA-MB-231 respectively, a transient 2.2- and 1.6-fold increase in intracellular
CCR7 MCF-7 ®lamentous actin (F-actin) in human breast cancer cells within 20 s
CCR6 MDA-MB-468 (Fig. 3a). Conversely, the chemokine CX3CL1/fractalkine, whose
CCR5 receptor CX3CR1/V28 was not detected on breast cancer cells
CCR4
(Fig. 1a), did not induce actin polymerization (Fig. 3a).
In tumour cells, high levels of actin polymerization are required
CCR3
for the formation of pseudopodia, which in turn are needed for the
CCR2 invasion of malignant cells into tissues and for ef®cient metastases
CCR1 formation18. Confocal laser scan microscopy of breast cancer cells
0 500 1,000 1,500 stimulated in suspension with either CXCL12 or CCL21 revealed
fg per 100 ng cDNA
intense F-actin staining in the periphery of the cells and a redistri-
bution of F-actin towards a leading edge (Fig. 3b±e). In adherent
breast cancer cells, distinct pseudopodia formation was observed after
20 min of stimulation with either CCL21 or CXCL12 (Fig. 3f, g).
b CX3CR1
In agreement with these ®ndings, both CXCL12 and CCL21
XCR1 induced directional migration of breast cancer cells (Fig. 4a±d)
CXCR5 and directional invasion through a reconstituted basement mem-
CXCR4
brane (Fig. 4e, f) in a dose-dependent manner. Optimal migratory/
invasive responses to CXCL12 (Fig. 4a, c, e) or CCL21 (Fig. 4b, d, f)
CXCR3 Melanocytes were observed at concentrations of 100 nM, or 100 and
CXCR2 TXM-40
200 nM, respectively, reminiscent of observations made with
UKRV-Mel-4
CXCR1 URKV-Mel-30 leukocytes16,19,20. Compared with breast cancer cells of well-char-
Malme-3M acterized cell lines (MDA-MB-231, MDA-MB-361), primary
CCR10
BLM tumour cells derived from a patient with malignant pleural effusion
HT-144
CCR9
SK-Mel-2
exhibited signi®cant chemotactic responses to both CXCL12 and
CCR8 MEWO CCL21 (Fig. 4c, d). CXCL12- and CCL21-mediated chemotaxis and
CCR7
MV3 invasion could be blocked by neutralizing anti-CXCR4 or anti-CCL21
SK-Mel-31 antibodies, respectively, con®rming the speci®city of the chemotactic
CCR6 SK-Mel-28
TXM-13
response induced by these chemokines (data not shown).
CCR5
CCR4 Migratory response to organ-derived proteins
To determine the biological relevance of CXCL12-mediated chemo-
CCR3
taxis in the context of all chemotactic factors present in the
CCR2 extracellular matrix of a particular organ, protein extracts of
CCR1 normal human lung, liver, skin and muscle, and conditioned
0 100 200 300 media from human primary bone-marrow and lymph-node stro-
mal cells, were tested for their chemotactic activity on breast cancer
fg per 100ng cDNA
cells (Fig. 5a±f). Protein extracts of lung and liver induced chemo-
Figure 1 Expression of chemokine receptors in human tumour cells. Quantitative RT± tactic responses in breast cancer cells, indicating that extracts
PCR analyses of all known chemokine receptors in seven human breast cancer cell lines derived from these organs contain factors with chemotactic proper-
compared with normal primary mammary epithelial cells (a), and in 12 human malignant ties (data not shown). Blocking the CXCL12/CXCR4 interactions
melanoma cell lines compared with normal primary melanocytes (b). Values are with neutralizing anti-CXCR4 (Fig. 5a, b) or anti-CXCL12 anti-
expressed as femtograms of target gene in 100 ng of total cDNA. bodies (see Supplementary Information) signi®cantly impaired

NATURE | VOL 410 | 1 MARCH 2001 | www.nature.com © 2001 Macmillan Magazines Ltd 51
articles
these migratory responses by 63±76% and 60±62%, respectively,
indicating that CXCL12 is one of the main chemotactic factors for
breast cancer cells in the extracellular matrix of these organs.
Furthermore, migration of breast cancer cells in response to
conditioned medium from human bone-marrow (Fig. 5c) or
lymph-node (Fig. 5d) stromal cells was signi®cantly reduced (53±
61% and 51±54%, respectively) after CXCR4 blocking. In contrast,
protein extracts from organs that represent rare targets of breast
cancer metastasis, such as skin (Fig. 5e) or muscle (Fig. 5f),
exhibited weak overall chemoattractive properties for breast
cancer cells, compared with protein extracts from lung and liver.
Notably, these migratory responses were not affected by CXCR4
neutralization (Fig. 5e, f).
CXCR4-neutralization inhibits metastasis in vivo
Having established that breast cancer cells express functionally
active CXCR4, we evaluated the contribution of CXCL12/CXCR4
Figure 2 Expression of CCR7 and CXCR4 in breast cancer. a, b, Quantitative RT±PCR
analyses of CCR7 (a) and CXCR4 (b) mRNA expression in primary breast carcinoma
(n = 12), normal mammary gland tissue (n = 5), primary mammary epithelial (n = 3) and
stromal (n = 3) cells. Mean 6 s.e.m. (Student's t-test; *P , 0.05, **P , 0.005).
c±n, Immunohistochemical evaluation of CXCR4 expression. c±e, Section of normal
mammary gland tissue. f±h, i±k, Invasive ductal carcinoma of two different patients.
52 © 2001 Macmillan Magazines Ltd
interactions on metastasis in vivo by using experimental and
spontaneous metastasis models of breast cancer. The CXCR4-
positive, human breast carcinoma cell line MDA-MB-231 was
injected either intravenously (i.v.) into the tail vein or orthotopi-
cally into the mammary fat pad of severe combined immunode®-
cient (SCID) mice.
In vitro, chemotaxis and invasion assays con®rmed that MDA-
MB-231 cells respond to murine CXCL12 (data not shown). Real-
time quantitative PCR analyses showed that there was signi®cant
expression of human CXCR4 mRNA in both primary tumours and
metastases-in®ltrated lungs of SCID mice that had been injected
with MDA-MB-231 cells (Fig. 6a). Immunohistochemical staining
for human CXCR4 depicted strong protein expression by tumour
cells of primary tumours (Fig. 6c, d) and lung metastases (Fig. 6e, f),
indicating that MDA-MB-231 cells maintained high levels of
CXCR4 expression in vivo. In agreement with CXCL12 mRNA
expression in normal human organs (Fig. 2o), quantitative analysis
l±n, Axillary lymph-node metastasis. Original magni®cation, ´200 (c, f, i, l, anti-CXCR4;
e, h, k, n, isotype); ´400 (d, g, j, m, anti-CXCR4). Quantitative RT±PCR analyses of
CXCL12/SDF-1 (o), CCL21/6Ckine and CCL19/MIP-3b (p) or CCL27/CTACK (q) mRNA
expression in various normal human organs. Values are expressed as femtograms of
target gene in 100 ng of total cDNA.
NATURE | VOL 410 | 1 MARCH 2001 | www.nature.com

of murine CXCL12 mRNA in normal organs of SCID mice showed
peak levels of expression in mouse lung, lymph nodes, bone marrow
and liver (data not shown).
Twenty-eight days after i.v. injection of MDA-MB-231 cells and
the twice weekly treatment with either neutralizing anti-human
CXCR4 monoclonal antibody or isotype control, a signi®cant
articles
decrease in lung metastasis was observed in anti-CXCR4-treated
mice (relative suppression, 61±68%; P , 0.001) (Fig. 7a±c, e).
Furthermore, spontaneous metastasis to the lung 44 days after
orthotopic injection of MDA-MB-231 cells was impaired signi®-
cantly by treatment with a neutralizing anti-CXCR4 monoclonal
antibody (relative suppression, 73±82%; P , 0.001) (Fig. 7d, f).
In addition to the histological evaluation of lung in®ltrates, the
individual tumour burden per lung was measured by quantitative
PCR using primers speci®c for the human housekeeping gene
HPRT. Expression of human HPRT was detectable in cultured
human MDA-MB-231 breast carcinoma cells, primary tumours
and tumour-in®ltrated lungs of untreated SCID mice, but unde-
tectable in normal mouse lungs (Fig. 6b). The human HPRT mRNA
content correlated directly with the tumour load of each sample.
Subsequent quantitative PCR analyses of human HPRT in lungs of
either isotype- or anti-CXCR4-treated mice indicated a signi®cant
reduction in lung-in®ltrating tumour cells after anti-CXCR4 treat-
ment in both experimental (i.v.) and spontaneous metastasis
models (relative suppression, 61±73%; P , 0.05) (Fig. 7e, f).
Furthermore, levels of human CXCR4 mRNA in the lungs of
CXCR4-treated mice showed a relative suppression, ranging
from 67 to 89%, as compared with isotype-treated mice (Fig. 7g,
h). Thus, quantitative PCR analyses con®rmed the histological
quanti®cation.
Clinical evaluation of draining lymph nodes 44 days after ortho-
topic injection of MDA-MB-231 cells showed that anti-CXCR4
treatment inhibited metastasis to inguinal and axillary lymph nodes

a b

Migrated cells per 5 fields

Migrated cells per 5 fields

500 6 500 4
CXCL12 CCL21

Chemotaxis index

Chemotaxis index
400 * *
400 3
* 4
300 300
2
200 200
2
1
100 100
0 0 0 0
100
200
0.1
1
10
0

100
200
400
0.1

10
0

1
nM nM
c d
Migrated cells per 5 fields

Migrated cells per 5 fields

400 CXCL12 2 200 2

CCL21
* * *

Chemotaxis index
Chemotaxis index

1.5 150 * 1.5

300

200 1 100 1

100 0.5 50 0.5

0 0 0 0
10
100
200
400
1
0.1
0

100
200
400
10
1
0

nM nM
e f
Invaded cells per 5 fields
Invaded cells per 5 fields

2,000 8 2,500 4
CXCL12 * CCL21
*
2,000
Invasion index

Invasion index

1,500 6 * 3
*
* * 1,500
1,000 4 2
1,000
500 2 1
500
0 0 0 0
1
0.1

100
200

100
200
400
0

0.1
10

10
1
0

nM nM

Figure 3 F-actin polymerization in human breast carcinoma cells. a, Intracellular F-actin
content measured by ¯ow cytometry in MDA-MB-231 cells after stimulation with CXCL12/
SDF-1a (100 nM), CCL21/6Ckine (100 nM) or CX3CL1/fraktalkine (100 nM). Data points
are plotted relative to the mean ¯uorescence before the addition of chemoattractant. b±e,
Confocal microscopy series of breast carcinoma cells after CXCL12 (100 nM) stimulation
in suspension. MDA-MB-231 cells unstimulated (b) or stimulated for 20 s (c), 3 min (d) or
20 min (e). f, g, Adherent MDA-MB-231 cells unstimulated (f) or stimulated (g) with
CXCL12 (100 nM) for 20 min. Scale bar, 50 mm.
Figure 4 Chemokine-mediated migration and invasion of breast carcinoma cells.
Chemotactic response of MDA-MB-231 cells to different concentrations of CXCL12/SDF-
1a (a) or CCL21/6Ckine (b). Directional migration of primary breast carcinoma cells
towards CXCL12 (c) or CCL21 (d) gradients. Effects of various concentrations of CXCL12
(e) or CCL21 (f) on the invasion of MDA-MB-361 cells. Results are expressed as the mean
number of migrating cells per well. Chemotaxis indices were calculated as ratio of cells
migrated toward a chemokine gradient to cells migrated in the negative control.
Mean 6 s.d. (Student's t-test; *P , 0.005).

NATURE | VOL 410 | 1 MARCH 2001 | www.nature.com © 2001 Macmillan Magazines Ltd 53
articles
(Table 1). All isotype-treated mice developed ipsi- and contralateral
inguinal, and axillary lymph-node metastases at the site of injection
(ipsilateral) with diameters exceeding 3 mm. In contrast, only 38%
of anti-CXCR4-treated mice presented inguinal lymph-node metas-
tases at the site of injection (diameter , 3 mm) (Table 1).
Discussion
It has been proposed that molecules regulating the metastatic
dissemination of tumour cells to speci®c anatomical sites need to
ful®l the following criteria1,2. First, they have to be constitutively
expressed at principal sites of metastasis. Second, adhesion of target
cells to the endothelium and transendothelial migration need to be
promoted. Third, these molecules must be capable of mediating the
invasion of cells into tissues that provide supportive microenviron-
ments. Last, this process requires the expression of a distinct
receptor repertoire by the target cells, depending on their metastatic
pro®le.
Given their well-established roles in leukocyte traf®cking and
homeostasis, chemokines are perfectly positioned to ful®l these
criteria4,5,7±10,21,22. We have shown that, out of all known chemokine
receptors, breast cancer cells speci®cally express functionally active
CXCR4 and CCR7, which trigger actin polymerization, pseudopo-
dia formation, and the directional migration and invasion of breast
cancer cells. Together with the distinct tissue distribution of their
ligands CXCL12 and CCL21 in organs representing the main sites of
breast cancer metastasis and the signi®cant inhibition of breast
cancer metastasis observed in vivo, this ®nding indicates that
chemokine ligand±receptor interactions may be crucial for the
metastatis of breast cancer.
Although chemokines have been implicated in tumour cell
growth2,23, angiogenesis24 and the host immune response against
malignant cells2,25, our study provides evidence that chemokines
may promote metastasis by acting directly on tumour cell migration
and invasion. Both CXCR4 and CCR7 are critical for cell traf®cking
and tissue homeostasis4,9,10,26±29. CXCL12 is the only known ligand
for CXCR4 (refs 26±29). This chemokine ligand±receptor pair

a b a b
30 20
MDA-MB-231 cells
Chemotaxis index
Chemotaxis index

15
20
Primary tumour 1
10 * 2
* 3
10 4
* 5 5
* 6
0 0 7
0 ct –l 1 –l 1 –1 –1 0 act –l 1 –l 1 –l 1 –l 1
l l
tra m g m g m g m tr m m m m
x
e µg µ µ µ ex µg µg µg µg Normal lung 1
n
ein 0 0 10 0 e i 1 0 20 0 0 2
ot 4 1 2 e e 2 ot 4 4 e 1 e 2
pr CR CR4 otyp typ r pr CR CR typ typ 3
n g X X Is so e X X o o 4
Lu ti-C ti-C + + I Liv nti-C ti-C + Is + Is
an an a n
+ + + +a Lung+metastases 1
c d 2
6 10 3
4
Chemotaxis index

Chemotaxis index

5
7.5 5
4 6
* 7
3 * 5 hCXCR4 hHPRT
* *
2
60

15
40

10
20

80

20
0
0

5
2.5
1
0 0 fg per 25 ng cDNA fg per 25 ng cDNA
0 M –l 1 –l 1 –1 –1 0 M –l 1 –l 1 –l 1 –l 1
l
C m m m ml C m m m m
w g
g µg µ µ g de µg g g g
µ no 10 20 µ 10 µ 20 µ
r o
ar 0 0 0 0 c d
m 41 42 e1 e2 ph 4 4
ne CR CR otyp typ m R R pe pe
o
B CX CX Is Iso Ly CXC XC soty soty
C
ti- ti- + + ti- ti- + I + I
an an an n
+ + + +a
e f
3
2
Chemotaxis index
Chemotaxis index

1.5 2

1 e f
1
0.5

0 0
0 ct –l 1 –l 1 0 ct –l 1 –l 1
tra m m tra m m
ex µg µg ex µg µg
e i n 0 0 i n 0 20
ot 4 1 2 e
ot 4 1 4
pr CR CR4 pr CR CR
in X X cle X CX
Sk ti-C ti-C us i-C i-
an an M ant ant
+ + + +
Figure 6 Expression of human CXCR4 and human HPRT in the MDA-MB-231 breast
Figure 5 Migration of breast carcinoma cells in response to organ-derived proteins. cancer metastasis model. Quantitative real-time RT±PCR (TaqMan) analyses of
Chemotaxis of MDA-MB-231 breast carcinoma cells in response to protein extracts of hCXCR4 (a) and hHPRT (b) mRNA expression in primary tumours (n = 7), metastases-
human lung (a) or liver (b), conditioned medium (CM) from human primary bone marrow in®ltrated lungs (n = 7), and normal lungs of SCID mice. Values are expressed as
(c) or lymph-node stromal cells (d) and protein extracts from human skin (e) or muscle (f) femtograms of target gene in 25 ng of total cDNA. Immunohistochemistry of human
in the presence of a neutralizing anti-CXCR4 monoclonal antibody or isotype control. CXCR4 in primary tumours (c, d) and lung metastases (e, f). Representative staining in
Chemotaxis indices were calculated as ratio of cells migrated toward a protein gradient to tissue samples from one out of six mice. Original magni®cation, ´400 (c, e, anti-CXCR4;
cells migrated in the negative control. Mean 6 s.d. (Student's t-test; *P , 0.001). d, f, isotype).

54 © 2001 Macmillan Magazines Ltd NATURE | VOL 410 | 1 MARCH 2001 | www.nature.com
 articles

a Table 1 Frequency of lymph-node metastases

b
Mice* Ipsilateral Contralateral Axillary LN
inguinal LN inguinal LN
Isotype (n = 8) 8/8 (.3 mm) 8/8 (.3 mm) 8/8 (.3 mm)
Anti-CXCR4 (n = 8) 3/8 (,3 mm) 0/8 0/8
.............................................................................................................................................................................
* MDA-MB-231 breast carcinoma cells (107) were injected into the inguinal mammary fat pad of
SCID mice, which were treated with anti-CXCR4 monoclonal antibody or isotype control. Regional
lymph-node (LN) metastasis was evaluated on day 44.

c 1 d 500
82%
68%

Metastases number
0.8 400
preferential expression of CXCL12 in lung, liver and lymph node
Metastasis index

also suggests a role for this chemokine in the tropism of breast

per lung

0.6 300

per lung
0.4
**
200
cancer cells for these anatomical sites.
**
Data point to the crucial role of CCL21 and its receptor CCR7 in
0.2 100 the homing of lymphocytes into secondary lymphoid organs. A
0 0 natural mutation in mice, designated plt (for paucity of lymph-node
Isotype anti-CXCR4 Isotype anti-CXCR4
(n=8) (n=8) (n=8) (n=8) T cells) that results in the loss of one of the forms of mCCL21 (refs
e 10 f 5 14, 15, 30), and targeted disruption of the CCR7 gene10 cause
61% impaired homing of naive T cells to secondary lymphoid organs.
hHPRT (fg) per lung

hHPRT (fg) per lung

4 73%
7.5 Thus, the abundant expression of the homeostatic chemokine
3 CCL21 in lymph nodes makes it a likely candidate to attract
5
* 2 * CCR7-positive tumour cells. In fact, CXCL12 and CCL21 may
2.5 have synergistic effects, as both chemokines are highly expressed
1
in lymph nodes and mediate chemotactic and invasive responses of
0
Isotype anti-CXCR4
0
Isotype anti-CXCR4
breast cancer cells in vitro.
(n=8) (n=8) (n=8) (n=8) Our ®ndings are probably not unique to breast cancer. Other
g 5 h 8 tumour entities of haematopoietic and non-haematopoietic origin,
hCXCR4 (fg) per lung

89%
including acute myeloid and lymphoblastic leukaemia31, chronic
hCXCR4 (fg) per lung

4 67%
6 lymphocytic leukaemia17,32, non-Hodgkin B-cell lymphoma33 and
3
4
2 that mediate tumour cell migration in vitro. Our results in breast
1
**
2 cancer and malignant melanoma suggest that malignant cells, in
0
Isotype anti-CXCR4
0
Isotype anti-CXCR4
(n=8) (n=8)
Figure 7 Effect of CXCR4-neutralization on metastasis in vivo. a, b, Haematoxylin and
eosin staining of lungs from SCID mice injected with human MDA-MB-231 breast
carcinoma cells and treated with either isotype control (a) or anti-hCXCR4 monoclonal
antibody (b). Original magni®cation, ´100. c, Lung colony formation after i.v. injection of
MDA-MB-231 cells in isotype- or anti-CXCR4-treated mice. Representative data of one
out of two experiments. d, Spontaneous lung metastasis after orthotopic injection of MDA-
MB-231 cells. e, f, Metastases quanti®cation by quantitative RT±PCR analyses in
contralateral lungs. Expression of hHPRT mRNA after i.v. (e) or orthotopic (f) injection.
Expression of hCXCR4 mRNA after i.v. (g) or orthotopic (h) injection. Values are expressed
as femtograms of target gene in 25 ng of total cDNA. Mean 6 s.e.m. (Student's t-test;
*P , 0.05, **P , 0.001). Percentages indicate relative suppression compared with
isotype.
exhibits unparalleled chemotactic ef®cacy for leukocytes in vitro
and is a highly potent chemoattractant in vivo16,19,26±29. Both
CXCR4- and CXCL12-de®cient mice display perinatal lethality
owing to profound defects in embryonic development of the
haematopoietic, cardiovascular and nervous systems26±29. These
phenotypic changes are mediated by the disrupted migration of
embryonic progenitor cells into appropriate microenvironments,
suggesting that CXCL12/CXCR4 interactions are vital for the
migration of haematopoietic and non-haematopoietic cells in vivo.
Furthermore, in vivo studies using neutralizing antibodies to
CXCR4 implicate this receptor in the homing and repopulation of
human stem cells into the bone marrow of SCID mice9. As expres-
sion of CXCR4 by breast cancer cells is related to ef®cient, CXCL12-
induced migration and invasion of these cells both in vitro and in
vivo, it is conceivable that CXCL12/CXCR4 interactions may con-
tribute to bone-marrow in®ltration by breast cancer cells. The
pancreatic cancer34, express functionally active chemokine receptors
**
general, express distinct and non-random patterns of chemokine
receptors. Furthermore, the association of CCR10 expression by
(n=8) (n=8) malignant melanoma cells with the skin-speci®c expression of its
ligand CCL27/CTACK7,8 and the high incidence of skin metastases
in this malignant disease support the involvement of chemokine
receptors in metastasis.
Currently, intense efforts are underway to identify small-mole-
cule antagonists for many chemokine receptors35. We propose that
small molecule antagonists of chemokine receptors, such as CXCR4,
may be useful to interfere with tumour progression and metastasis
in tumour patients. M
Methods
Cell lines
The following human breast cancer and melanoma cell lines were obtained from the
ATCC: DU-4475, MCF-7, T-47D, Hs578T, MDA-MB-468, MDA-MB-361, MDA-MB-231,
SK-Mel-2, SK-Mel-28, SK-Mel-31, HT-144 and Malme-3M. MV3 melanoma cells were a
gift from D. J. Ruiter, Nijmegen; TXM-14 and TXM-30 cells were obtained from I. J. Fidler,
Houston; UKRV-Mel-4 and UKRV-Mel-30 melanoma cells were a gift from
D. Schadendorf, Mannheim. Human primary mammary epithelial and stromal cells, and
human primary melanocytes and bone-marrow stromal cells were obtained from
Clonetics (San Diego, CA). Human primary lymph-node stromal cells were cultured from
normal human lymph-node tissue as described33.
Real-time quantitative PCR
Total RNA from cells or homogenized tissue samples was extracted and reverse transcribed
as described8. Complementary DNA was quantitatively analysed for the expression of
human chemokines and chemokine receptors by the ¯uorogenic 59-nuclease PCR assay36
as reported8. Speci®c primers and probes were obtained from Applied Biosystems. Gene-
speci®c PCR products were continuously measured by means of an ABI PRISM 7700
Sequence Detection System (Applied Biosystems) during 40 cycles. For metastasis
quanti®cation, speci®c primers for human HPRT which do not cross-react with its mouse
counterpart were designed (forward, 59-TTCCTTGGTCAGGCAGTATAATCC-39; reverse,
59-AGTCTGGCTTATATCCAACACTTCG-39). 18S ribosomal RNA was used for
normalization.
Tissue samples and immunohistochemistry
Tissue samples of primary tumours from untreated patients (n = 12) with invasive ductal

NATURE | VOL 410 | 1 MARCH 2001 | www.nature.com © 2001 Macmillan Magazines Ltd 55
articles
(n = 9) or lobular (n = 3) breast carcinoma were taken, with informed consent, from
either diagnostic biopsies or after lumpectomy/mastectomy. Histopathological diag-
nosis was con®rmed for each specimen. Tissue samples of normal human organs were
obtained from the National Disease Research Interchange. RNA of various normal
human organs was obtained from Clontech. The study was approved by the local ethics
committee.
Primary tumours and metastases were routinely ®xed with formalin and embedded in
paraf®n. Antigen retrieval was performed and sections were stained with either anti-
human CXCR4 monoclonal antibody (12G5, IgG2a) or IgG2a isotype (R&D Systems) using
a standard indirect avidin±biotin horseradish peroxidase method. Colour was developed
with diaminobenzidine (DAB) and sections were counterstained with haematoxylin. For
staining human CXCR4 in paraf®n-embedded or cryo-preserved tissue samples of
tumour-bearing SCID mice, 3-amino-9-ethylcarbazole (AEC) was used for colour
development.
Actin polymerization assay and microscopy
Actin polymerization was tested as described17,37. Breast cancer cells were incubated either
with CXCL12/SDF-1a or CCL21/6Ckine (R&D Systems) at concentrations shown to be
optimal in chemotaxis experiments, or with CX3CL1/fractalkine as negative control. At the
indicated time points, cells were ®xed, permeabilized and stained in a solution containing
paraformaldehyde, lysophosphatidylcholine and ¯uorescein isothiocyanate (FITC)-
labelled phalloidin. Fixed cells were subjected to ¯ow cytometry or analysed by confocal
microscopy (Leica DMR). For experiments with adherent cells, breast cancer cells were
pre-seeded and incubated with CXCL12, CCL21 or assay buffer (DMEM/0.1% bovine
serum albumin (BSA)/ 12 mM HEPES) for 20 min. Cells were stained with phalloidin and
49,6-diamidino-2-phenylindole dihydrochloride (DAPI) (Molecular Probes).
Chemotaxis and chemoinvasion assays
Migration and invasion was assayed in 24-well cell-culture chambers using inserts with
8-mm pore membranes as described38. Membranes were pre-coated with ®bronectin (2.5±
7.5 mg ml-1) for chemotaxis or Matrigel (28 mg per insert) and ®bronectin for invasion
studies. Breast cancer cells were resuspended in chemotaxis buffer (DMEM/ 0.1% BSA/
12 mM HEPES) at 2 or 4 ´ 104 cells per ml. After incubation for 6 or 24 h for chemotaxis or
chemoinvasion assays, respectively, cells on the lower surface of the membrane were
stained and counted under a light microscope in at least ®ve different ®elds (original
magni®cation, ´200). Assays were performed in triplicates. Chemokinesis was tested in
checkerboard assays and was uniformly negative for both CXCL12 and CCL21.
Proteins from homogenized normal human lung, liver, skin and muscle were extracted in
Tris-HCl and protease inhibitor as described39. Conditioned media from human primary
bone-marrow or lymph-node stromal cells were generated as described16,17,33.
For neutralization studies, cells were pre-incubated with various concentrations of
either anti-human CXCR4 monoclonal antibody (44717.111, IgG2b), anti-human
CXCL12 polyclonal antibody (goat IgG) or anti-human CCL21 polyclonal antibody (goat
IgG) (all R&D Systems).
In vivo metastasis studies
CB-17 SCID mice (Taconic Farms, Germantown, NY) were injected with MDA-MB-231
breast carcinoma cells either i.v. (106 cells) into the tail vein, orthotopically into the
inguinal mammary fat pad (107 cells). Mice were treated with intraperitoneal injections of
a neutralizing anti-human CXCR4 antibody (44717.111, IgG2b) or control IgG2b twice
weekly (1 mg per injection). No cytotoxicity for either the anti-human CXCR4 antibody or
control IgG2b could be detected in in vitro assays. For studies of experimental metastasis,
lungs were collected on day 28 and ®xed or snap frozen for immunohistochemistry and
RNA extraction.
Micro-metastasis was quanti®ed by counting the total tissue area per lung section (D1)
and micro-metastasis present in the same area (D2) using a 21-mm2 reference grid. The
metastatic index was calculated by the ratio D2/D1. For spontaneous metastasis experi-
ments, primary tumours and organs were collected on day 44 after injection. The number
of lung colonies was counted by light microscopy (magni®cation, ´40). The anti-human
CXCR4 monoclonal antibody showed no cross-reactivity with murine CXCR4.
Received 15 May 2000; accepted 17 January 2001.
1. Nicolson, G. L. Paracrine and autocrine growth mechanisms in tumor metastasis to speci®c sites
with particular emphasis on brain and lung metastasis. Cancer Metastasis Rev. 12, 325±343
(1993).
2. Wang, J. M., Deng, X., Gong, W. & Su, S. Chemokines and their role in tumor growth and metastasis.
J. Immunol. Methods 220, 1±17 (1998).
3. Yeatman, T. J. & Nicolson, G. L. Molecular basis of tumor progression: mechanisms of organ-speci®c
tumor metastasis. Semin. Surg. Oncol. 9, 256±263 (1993).
4. Zlotnik, A. & Yoshie, O. Chemokines: a new classi®cation system and their role in immunity.
Immunity 12, 121±127 (2000).
5. Campbell, J. J. & Butcher, E. C. Chemokines in tissue-speci®c and microenvironment-speci®c
lymphocyte homing. Curr. Opin. Immunol. 12, 336±341 (2000).
6. Butcher, E. C., Williams, M., Youngman, K., Rott, L. & Briskin, M. Lymphocyte traf®cking and
regional immunity. Adv. Immunol. 72, 209±253 (1999).
7. Morales, J. et al. CTACK, a skin-associated chemokine that preferentially attracts skin-homing
memory T cells. Proc. Natl Acad. Sci. USA 96, 14470±14475 (1999).
8. Homey, B. et al. Cutting edge: the orphan chemokine receptor G protein-coupled receptor-2 (GPR-2,
CCR10) binds the skin-associated chemokine CCL27 (CTACK/ALP/ILC). J. Immunol. 164, 3465±
3470 (2000).
9. Peled, A. et al. Dependence of human stem cell engraftment and repopulation of NOD/SCID mice on
56 © 2001 Macmillan Magazines Ltd
CXCR4. Science 283, 845±848 (1999).
10. Forster, R. et al. CCR7 coordinates the primary immune response by establishing functional
microenvironments in secondary lymphoid organs. Cell 99, 23±33 (1999).
11. Beahrs, O. H. in Manual for Staging of Cancer 4th edn (JB Lippincott, Philadelphia, 1992).
12. Tashiro, K. et al. Signal sequence trap: a cloning strategy for secreted proteins and type I membrane
proteins. Science 261, 600±603 (1993).
13. Bruce, J., Carter, D. C. & Fraser, J. Patterns of recurrent disease in breast cancer. Lancet 1, 433±435
(1970).
14. Gunn, M. D. et al. A chemokine expressed in lymphoid high endothelial venules promotes the
adhesion and chemotaxis of naive T lymphocytes. Proc. Natl Acad. Sci. USA 95, 258±263 (1998).
15. Gunn, M. D. et al. Mice lacking expression of secondary lymphoid organ chemokine have defects in
lymphocyte homing and dendritic cell localization. J. Exp. Med. 189, 451±460 (1999).
16. Bleul, C. C., Fuhlbrigge, R. C., Casasnovas, J. M., Aiuti, A. & Springer, T. A. A highly ef®cacious
lymphocyte chemoattractant, stromal cell-derived factor 1 (SDF-1). J. Exp. Med. 184, 1101±1109
(1996).
17. Burger, J. A., Burger, M. & Kipps, T. J. Chronic lymphocytic leukemia B cells express functional
CXCR4 chemokine receptors that mediate spontaneous migration beneath bone marrow stromal
cells. Blood 94, 3658±3667 (1999).
18. Verschueren, H., Van der Taelen, I., Dewit, J., De Braekeleer, J. & De Baetselier, P. Metastatic
competence of BW5147 T-lymphoma cell lines is correlated with in vitro invasiveness, motility and
F-actin content. J. Leukoc. Biol. 55, 552±556 (1994).
19. Bleul, C. C. et al. The lymphocyte chemoattractant SDF-1 is a ligand for LESTR/fusin and blocks
HIV-1 entry. Nature 382, 829±833 (1996).
20. Hromas, R. et al. Isolation and characterization of Exodus-2, a novel C-C chemokine with a unique
37-amino acid carboxyl-terminal extension. J. Immunol. 159, 2554±2558 (1997).
21. Hedrick, J. A. & Zlotnik, A. Identi®cation and characterization of a novel beta chemokine containing
six conserved cysteines. J. Immunol. 159, 1589±1593 (1997).
22. Campbell, J. J. et al. Chemokines and the arrest of lymphocytes rolling under ¯ow conditions. Science
279, 381±384 (1998).
23. Luboshits, G. et al. Elevated expression of the CC chemokine regulated on activation, normal T
cell expressed and secreted (RANTES) in advanced breast carcinoma. Cancer Res. 59, 4681±4687
(1999).
24. Belperio, J. A. et al. CXC chemokines in angiogenesis. J. Leukoc. Biol. 68, 1±8 (2000).
25. Vicari, A. P. et al. Antitumor effects of the mouse chemokine 6Ckine/SLC through angiostatic and
immunological mechanisms. J. Immunol. 165, 1992±2000 (2000).
26. Ma, Q. et al. Impaired B-lymphopoiesis, myelopoiesis, and derailed cerebellar neuron migration in
CXCR4- and SDF-1-de®cient mice. Proc. Natl Acad. Sci. USA 95, 9448±9453 (1998).
27. Tachibana, K. et al. The chemokine receptor CXCR4 is essential for vascularization of the gastro-
intestinal tract. Nature 393, 591±594 (1998).
28. Nagasawa, T. et al. Defects of B-cell lymphopoiesis and bone-marrow myelopoiesis in mice lacking the
CXC chemokine PBSF/SDF-1. Nature 382, 635±638 (1996).
29. Zou, Y. R., Kottmann, A. H., Kuroda, M., Taniuchi, I. & Littman, D. R. Function of the
chemokine receptor CXCR4 in haematopoiesis and in cerebellar development. Nature 393, 595±
599 (1998).
30. Vassileva, G. et al. The reduced expression of 6Ckine in the plt mouse results from the deletion of one
of two 6Ckine genes. J. Exp. Med. 190, 1183±1188 (1999).
31. Mohle, R. et al. Functional response of leukaemic blasts to stromal cell-derived factor-1 correlates with
preferential expression of the chemokine receptor CXCR4 in acute myelomonocytic and lympho-
blastic leukaemia. Br. J. Haematol. 110, 563±572 (2000).
32. Mohle, R., Failenschmid, C., Bautz, F. & Kanz, L. Overexpression of the chemokine receptor CXCR4 in
B cell chronic lymphocytic leukemia is associated with increased functional response to stromal cell-
derived factor-1 (SDF-1). Leukemia 13, 1954±1959 (1999).
33. Arai, J., Yasukawa, M., Yakushijin, Y., Miyazaki, T. & Fujita, S. Stromal cells in lymph nodes attract
B-lymphoma cells via production of stromal cell-derived factor-1. Eur. J. Haematol. 64, 323±332
(2000).
34. Koshiba, T. et al. Expression of stromal cell-derived factor 1 and CXCR4 ligand receptor system in
pancreatic cancer: a possible role for tumor progression. Clin. Cancer Res. 6, 3530±3535 (2000).
35. Hendrix, C. W. et al. Pharmacokinetics and safety of AMD-3100, a novel antagonist of the CXCR-4
chemokine receptor, in human volunteers. Antimicrob. Agents Chemother. 44, 1667±1673 (2000).
36. Holland, P. M., Abramson, R. D., Watson, R. & Gelfand, D. H. Detection of speci®c polymerase chain
reaction product by utilizing the 59!39 exonuclease activity of Thermus aquaticus DNA polymerase.
Proc. Natl Acad. Sci. USA 88, 7276±7280 (1991).
37. Howard, T. H. & Meyer, W. H. Chemotactic peptide modulation of actin assembly and locomotion in
neutrophils. J. Cell Biol. 98, 1265±1271 (1984).
38. Saiki, I., Murata, J., Nakajima, M., Tokura, S. & Azuma, I. Inhibition by sulfated chitin derivatives of
invasion through extracellular matrix and enzymatic degradation by metastatic melanoma cells.
Cancer Res. 50, 3631±3637 (1990).
39. Hujanen, E. S. & Terranova, V. P. Migration of tumor cells to organ-derived chemoattractants. Cancer
Res. 45, 3517±3521 (1985).
Supplementary information is available on Nature's World-Wide web site
(http://www.nature.com) or as paper copy from the London editorial of®ce of Nature.
Acknowledgements
We thank R. M. Barcenas for technical assistance; S. Ulloa and K. Smith for help
procuring tissue samples; and D. J. Ruiter, I. J. Fidler and D. Schadendorf for the
melanoma cell lines. We are grateful to R. de Waal Malefyt for establishing the quantitative
RT±PCR at the DNAX Research Institute, and H. Kanzler, T. Hauser and L. Bakker for
critical comments on the manuscript. B.H. was supported by a grant from the Deutsche
Forschungsgemeinschaft. DNAX Research Institute is supported by the Schering-Plough
Corporation.
Correspondence and requests for materials should be addressed to A.Z.
(e-mail: azlotnik@eosbiotech.com).
NATURE | VOL 410 | 1 MARCH 2001 | www.nature.com